Deal screening is the triage step that decides which commercial real estate opportunities are worth a full underwrite and which get set aside. It is not the buy decision, and it is not underwriting. It is the fast first pass where you compare an incoming deal against your firm’s criteria, read the summary numbers in the offering memorandum without verifying every line, and answer one question: is this worth my analyst’s next four hours? Most investment firms lose more good deals here than anywhere else, because screening is where a busy team decides what to ignore. This piece explains what screening is, how the funnel works, what a buy box does, and where AI helps a lean firm screen more deals without pretending to replace judgment.
Deal screening, defined
Screening is the process of evaluating many deals against a defined set of investment criteria to find the few that deserve deeper work. A broker sends an offering memorandum, or a deal shows up on CoStar or Crexi, and someone gives it a few minutes: does the asset type fit, is it in a market we buy, does the asking price put the return near our range? If yes, it advances. If no, it stops. That yes-or-no, repeated across every deal that hits your inbox, is screening.
The important word is reasonableness. When you screen, you are not checking whether the seller’s proforma is honest. You are checking whether the deal is in the neighborhood of something you would buy. You read the cap rate the broker printed, the in-place rents, the price per unit or per square foot, and you decide whether those numbers are plausible enough to justify pulling the real documents. Screening trusts the summary; underwriting verifies it.
For a small firm, screening is also the most neglected part of the pipeline. Sourcing gets attention because it feels like growth. Underwriting gets attention because it is where the math lives. Screening happens in the cracks, on a phone between property tours, and that is precisely why it leaks.
The screening funnel: why volume at the top wins
The standard mental model is a funnel. Deals pour in at the top, pass through a series of filters, and only the strongest survive to the bottom, where they get a full underwrite and maybe an offer. A firm buying ten assets a year might screen several hundred to get there. The ratio is supposed to be brutal; that is the point. Screening kills deals cheaply so your team spends its expensive hours only on the survivors. Two things follow that lean firms often miss.
First, the quality of what comes out the bottom is capped by the volume that goes in the top. If your firm only ever looks at the twenty deals a month that land in the principal’s inbox, your acquisitions are drawn from a puddle. Firms that scale deal flow deliberately, by staying visible to more brokers and monitoring more listing sources, get a better bottom of the funnel without changing their criteria at all. Deal-management vendors like Dealpath build their whole pitch around this: widen the intake, then filter faster.
Second, and this matters most for a four-to-twenty-person shop, your screening capacity is a fixed pipe. A team of three can only give real attention to so many deals a week by hand. Above that line, deals get triaged on the subject line or the broker’s reputation, and some of them are good. The cost of manual screening is not the hours your team spends; it is the deals that arrive above your capacity and never get opened. We priced that hidden cost in our breakdown of what screening deals by hand actually costs a lean firm: the labor line is the smallest of five.
The buy box: the criteria that do the filtering
A buy box is the written-down version of what your firm buys. It is the filter the funnel runs on, and if it lives only in the principal’s head, screening cannot be delegated or automated. Writing it down is the most valuable thing a lean firm can do before touching any tool.
A useful buy box is specific enough that a deal can be scored against it in under two minutes. Here is a worked example for a hypothetical value-add multifamily buyer:
| Criterion | Buy box |
|---|---|
| Asset type | Multifamily, 1980s–2000s vintage, garden or mid-rise |
| Markets | Nashville, Charlotte, Raleigh, Tampa metros |
| Size | 80–250 units |
| Price | $10M–$40M |
| Going-in cap rate | ≥ 5.5% on in-place NOI |
| Business plan | Light-to-moderate renovation, no ground-up |
| Occupancy | ≥ 85% at acquisition |
| Deal-breakers | Flood zone, single-employer town, deferred maintenance over ~$15K/unit |
With a box like this, screening becomes mechanical: check the asset type and market first, because those disqualify the fastest, then size and price, then the going-in return, then scan for deal-breakers. A deal that clears every row earns a full read; a deal that fails the market row stops on line one. That clarity is what makes the process fast, consistent, and eventually something a tool can help with. A vague buy box (“good deals in growing markets”) cannot be applied consistently by a tired human at 9 p.m., let alone by software. Notice too that most of the box is qualitative gating and only a few rows are numeric. Screening is mostly about fit, not math. The math comes later.
Screening vs. underwriting: the line lean firms blur
Screening ranks deals so you know what to open. Underwriting models the cash flows so you can price an offer. Screening answers “is this worth my time.” Underwriting answers “what is it worth.” They are different jobs with different tools, and vendors sell them interchangeably, which leads firms to buy a heavy modeling platform when their real bottleneck is triage.
The distinction has a practical edge. During screening you accept the broker’s numbers at face value on purpose, because verifying them is expensive and most deals will not survive the fit test anyway. During underwriting you distrust every number, re-derive the T-12, and check the comps. If you re-underwrite deals during screening, your funnel is too slow. If you skip verification during underwriting because “it looked good at screening,” you are about to overpay. Keeping the two stages separate is most of the discipline. Our field guide to the rules of AI-assisted underwriting covers the second half of that workflow.
Where AI belongs in screening, and where it must not
AI is genuinely useful in screening, in a narrow way. A general assistant like ChatGPT, Claude, or Gemini, or a purpose-built proptech tool, can read an offering memorandum, pull the asset type, unit count, in-place rents, asking price, and stated cap rate, compare them against your written buy box, and hand you a one-paragraph verdict with the key figures. Work that took twenty minutes of opening a PDF and retyping numbers can drop to a couple of minutes of review. For a firm processing hundreds of deals a year, that directly raises the throughput ceiling.
Here is the line. AI in screening should extract, rank, and summarize. It should not verify or decide. It has no proprietary transaction data, so it cannot tell you whether the seller’s rents are actually market. It states a wrong number with the same confidence as a right one, so a figure pulled from the wrong column looks identical to a correct one. And it cannot own the go/no-go, because that call carries capital and reputation that belong to a person. Used as an extraction layer, AI makes a lean team faster; used as a decision engine, it manufactures confident errors.
The vendor market blurs this constantly. You will see claims like “evaluate 100 deals in minutes” or accuracy figures north of 95%. Treat those as marketing until you have tested them on your own deals, and keep a human verifying the handful of numbers that would price an offer. Let AI narrow the pile and let a person decide what survives. We walk through building that intake step in lessons from automating deal intake for an acquisitions team, including where it needed a human gate.
One caution the tool reviews skip: deal packages are confidential, often under NDA. Before a principal at an IT-free firm uploads an OM to any assistant, they need three answers in writing: where the data is stored, whether uploads train shared models, and how history gets deleted. The contract is your only real safeguard, so vague answers are a reason to walk. This sits inside the broader question of adopting AI without an IT department, which we treat in the small-firm AI playbook.
What screening looks like without a system
Most 4-to-20-person firms screen with no system at all, and it works until it doesn’t. A broker’s email arrives, the principal skims the OM on a phone, forms a gut read, and either forwards it to an analyst or lets it slide down the inbox. There is no written buy box, no record of what was passed on, and no way to tell in December how many deals were seen versus seriously considered. The firm is screening; it just cannot see itself do it.
The failure mode is quiet. Good deals die because they arrived during a busy week. The same deal-breaker gets caught on Tuesday and missed on Friday because the person is tired. A written buy box and a consistent first-pass routine fix most of it before any software enters the picture, and they are prerequisites for AI helping at all. If the criteria are not explicit, no tool can apply them. For how screening connects to underwriting and market analysis across a lean acquisitions team, our deal analysis playbook is the place to start.
The firms that out-screen bigger competitors are not the ones with the fanciest tools. They are the ones who wrote down what they buy, kept the top of the funnel full, and drew a hard line between the fast triage that trusts the summary and the slow underwrite that trusts nothing. Start there, and the tools have something to stand on.
FAQ
What is deal screening in commercial real estate?
Deal screening is the first-pass evaluation of an incoming deal against your firm’s investment criteria to decide whether it is worth a full underwrite. You compare the asset type, market, size, price, and stated return in the offering memorandum against your buy box, without verifying every number, and either advance the deal or stop it. It is triage, not a buy decision. The goal is to kill deals cheaply so your team spends its expensive hours only on the ones that fit.
What is the difference between deal screening and underwriting?
Screening ranks deals so you know what to open; underwriting models the cash flows so you can price an offer. Screening trusts the summary numbers on purpose and asks “is this worth my time.” Underwriting distrusts every number, re-derives the financials, and asks “what is it worth.” A lean firm usually needs the screening layer first, because triage, not modeling, is what caps how many deals reach a real underwrite. Buying a heavy modeling platform when the real bottleneck is triage is a common and costly mismatch.
What is a buy box, and why does it matter for screening?
A buy box is the written-down set of criteria that defines what your firm buys: asset type, markets, size, price range, target return, business plan, and deal-breakers. It matters because it is the filter your screening funnel runs on. If the buy box lives only in the principal’s head, screening cannot be delegated, kept consistent, or automated. Writing it down so a deal can be scored against it in under two minutes is the most valuable step a firm can take before adopting any tool.
How many deals do firms screen to close one?
The ratio varies by strategy and market, but it is intentionally steep. A firm may screen several hundred deals to close a handful in a year, because screening exists to eliminate most opportunities quickly. The exact number matters less than the principle: the quality of what you close is capped by the volume you screen, so a firm drawing only from the deals that land in one inbox is limiting its outcomes before screening even begins.
Can AI screen commercial real estate deals?
AI can screen, in a specific sense: it can read an offering memorandum, extract the key figures, compare them to your buy box, and draft a ranked first-pass summary in a fraction of the manual time. What it cannot do is verify whether the numbers are true or make the go/no-go call, because it has no proprietary transaction data and will state a wrong figure as confidently as a right one. Use AI to extract, rank, and summarize; keep verification and the decision with a person.
Is it safe to upload deal packages to an AI tool?
Only if the contract allows it. Offering memorandums are confidential and frequently under NDA, so before uploading anything, get three answers in writing: where your data is stored, whether your uploads train shared models, and how you delete and export your history. “Your data is secure” is a marketing line; a data-processing agreement that commits the vendor not to train on your data and to delete on request is enforceable, and a firm without an IT department relies on that contract as its main safeguard.
How much does it cost to improve deal screening with AI?
It ranges from a modest training investment to a custom build. Getting your team fluent enough to run deal packages through a general assistant themselves is workshop-scale, with market rates for that kind of AI fluency training running roughly $2,000 to $15,000, and it often reveals whether you need anything more. A purpose-built screening workflow that integrates with your data sources is a larger project, typically tens of thousands to low six figures depending on scope. Price your current process first, so you are comparing against a real number.
Do I still need an analyst if AI screens my deals?
Yes; the analyst’s role shifts from extraction to judgment. The tool reads packages, ranks them against the buy box, and drafts summaries, removing the mechanical work of opening PDFs and retyping numbers. The person then verifies the figures that matter and decides what to underwrite. A lean team’s talent was never in transcribing rent rolls; it was in the go/no-go, the negotiation, and the broker relationships. The risk to manage is complacency, because a good tool works well enough that people stop checking, so build the verification step in on purpose.
Key takeaways
- Deal screening is triage: a fast first pass that decides which deals earn a full underwrite, based on fit against your criteria, not on verified numbers.
- The quality of what you close is capped by the volume you screen. A lean firm’s fixed screening capacity is the real constraint, and the deals it never opens are the hidden cost.
- A written buy box is the prerequisite for consistent, delegable, and automatable screening. If the criteria live only in someone’s head, no tool can apply them.
- Keep screening and underwriting separate: screening trusts the summary to move fast; underwriting trusts nothing to price an offer.
- AI belongs in screening as an extract-rank-summarize layer that raises throughput. It must not verify numbers or own the decision, and deal packages go into a tool only after the confidentiality terms are in writing.
Wondering where AI actually fits in your firm’s deal flow, and where it would introduce more risk than it removes? A short, free assessment maps that against your real deal volume, property types, and buy box faster than any tool comparison. Book your free AI-readiness assessment →
Arthur Wandzel
